Quantitative assessment of radiographic control informativeness using ROC analysis

2019 
In industrial radiography the issue of replacment film detectors by digital ones becomes increasingly relevant. Also remains unresolved the question of an objective comparison of various radiographic testing (RT). This article presents the results of research, aimed at obtaining a quantitative assessment of RT technology informativeness. In order to construct ROC curves, characterizing the quality of binary classification, using a specific RT technology, there was designed and manufactured a sample test, containing a simulation of the most difficult to detect discontinuities. Such test sample was made using additive technology. Developed manufacturing technology made it possible to ensure sufficient accuracy of linear dimensions, taking into account the product's design features. After exposure of test object to various detectors, there was a decoding procedure, conducted by experts, whose task was to establish the presence / absence of a defect in analyzed element of the image. Applying such technique, used in results analysis of deciphering the images of sample test, made it possible to minimize the influence of human factor and obtain ROC curves, reflecting the capabilities of RT technology only. The subsequent determination of ROC curves parameters allows to conduct a comparative analysis of informativeness of the technologies under consideration.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    4
    References
    1
    Citations
    NaN
    KQI
    []